RIPE IPmap - ENOG · 2018-06-11 · •Peering DB data and BGP data are used to reduce the locations probed •Score based on RTT, only RTT

[email protected] ENOG15 | 2018

Victor Naumov R&D RIPE NCC

RIPE IPmap Geolocating Internet Infra-Structure with Inference Engines and Crowdsourcing

But now for something completely different first

[email protected] ENOG15 | 2018 !3

User to User Still Important• The quality of end-user connections are often

expressed and optimised in download speeds towards content providers.

• Instead, these sketches focus on peer-to-peer connections with in a country.

• Fabric snapshots at points in time

• Trying to estimate an amount of ways the networks interconnect their users • Any “single point of failure”?

• How much IXPs and Transit Providers are involved?

We call it sketches since it uses rough estimates data from many source

[email protected] ENOG15 | 2018 !4

Data Sources: RIPE Atlas

• Measure the Internet! • By community, for community

• Wikipedia: https://en.wikipedia.org/wiki/RIPE_Atlas

[email protected] ENOG15 | 2018 !5

Data Sources: RIPE Atlas

10,500+ probes connected (310+ Anchors)

6000 results collected per second

40,000+ user-defined measurements weekly

Five types of user-defined measurements available to probe hosts and RIPE NCC members: ping, traceroute, DNS, SSL, NTP, HTTP, WiFi

[email protected] ENOG15 | 2018 !6

Data sources

• APNIC end-users per network estimation • https://stats.labs.apnic.net/aspop/

• IXP-country-jedi • https://www.ripe.net/ixp-country-jedi/

• mesh traceroutes between RIPE Atlas probes in a country

• measures and analyses (using IPMap)

[email protected] ENOG15 | 2018 !7

Data sources• CAIDA AS-to-ORG

• https://www.caida.org/data/as-organizations/

• PeeringDB • https://www.peeringdb.com/

• RIPE IPmap • https://ipmap.ripe.net/

• RIPEStat • https://stat.ripe.net/

Show me the Sketch!

[email protected] ENOG15 | 2018

The Netherlands

IXP

Transit�provider

size�of�a�segment�is�proportional��to�amount�of�end-users�

�in�the�networkthe�size�of�the�circle�indicates�the�in-

betweenness�of�the�network

end-users�network

Empty�segment�represents�ASes�providing�connectivity�to�<�1%�users

end-users�network�that�also�provides�transit

User to user without IXP can also be not optimized (the network is otpimised for using to services)

Transit proviuders are just in the middle of a user to user and they are not ISPs

in-betweennes is proportional to the number of traceroutes passing though the network

[email protected] ENOG15 | 2018

Russian Federation

No�peer-to-peer�data�in�the�AS�no�probes�present?

Orange�lines�-�end-users�networks��connected�through�an�IXP

green�lines�-�directly�connected

blue�lines�-�through�a�transit�network

dashed�lines�-�we�cannot�fully�map�this�path

dashed lines - where dns lookup is wrong or have stars in the trace route outputs

red lines show paths between IXPs

Why Geolocation of infra-structure?

[email protected] ENOG15 | 2018 12

Reasons• Increased interest in IP geolocation

• Content providers • From operators and researchers • The “geoloc” attribute is not a solution

• A unified geographical data format is needed

• IP geolocation is extremely difficult • Various approaches, some of them cannot be used singularly

• Academia is working on it! Let’s work together

• A validation/feedback loop is needed

• A unified geographical data format is needed

FOCUS ON

Infrastructure

[email protected] ENOG15 | 2018 13

Different Research Approaches• Triangulation a.k.a. trilateration

• Paristech anycast

• RIPE Atlas

• Reverse DNS based location inference • CAIDA DNS decoding database - DDec

• ‘Administrative’ analyses • PeeringDB

• RIPE Database

• Verification/falsification procedures

[email protected] ENOG15 | 2018 14

Commercial Offerings

• Tend to concentrate on the end-user IP addresses

• Opaque methodology

• IPv6 address space largely ignored

[email protected] ENOG15 | 2018 15

Our Integration attempt

• Accumulate research efforts as Inference Engines

• Each engine is applicable only in some cases

• Each engine has a score factor

• Complete transparency in inference methodology

[email protected] ENOG15 | 2018 16

Inference Engines

[email protected] ENOG15 | 2018 17

What is ipmap.ripe.net• A web application where you can query/

correct the geolocation of an IP address

• An API where you can query for multiple IP addresses in bulk

• An API where you can correct/provide yourself the geolocation of an IP address

• A web application to visualize traceroutes geolocated on a map

Massimo Candela | LACNIC 29 | May 2018 �18

Where is RIPE IPmap?

21

[email protected] ENOG15 | 2018

Where is RIPE IPmap?

19

here!

�19

[email protected] ENOG15 | 2018 20

IPmap Demo

[email protected] ENOG15 | 2018 21

IPmap Demo

[email protected] ENOG15 | 2018 22

IPmap Demo

Geolocation API

https://ipmap.ripe.net/api/v1

[email protected] ENOG15 | 2018 !24

/locatehttps://openipmap.ripe.net/api/v1/locate/83.163.50.165/best

*queries can be bundled with:

https://ipmap.ripe.net/api/v1/locate/all?resources=ip1,ip2,ip3..

[email protected] ENOG15 | 2018 !25

/locatehttps://openipmap.ripe.net/api/v1/locate/83.163.50.165/partials

[email protected] ENOG15 | 2018 !26

/locate - Active geolocation• If the IP has not been measured yet, a new Ping measurement starts

• Peering DB data and BGP data are used to reduce the locations probed

• Score based on RTT, only RTT <10ms are considered

• PeeringDB facilities and population bust the score

• A list of possible locations is returned

• We are working on it! (Contributions are welcome!)

[email protected] ENOG15 | 2018 !27

That’s why you need…

RIPE Atlas coverage!

What’s new?Short RIPE Atlas update

[email protected] ENOG15 | 2018 !29

Some New Features and Other Info

• Measurement tagging/labeling

• Result archives

• RIPE Atlas (storage) timestamps

• DNS-over-TLS support (API only yet)

• Scaling up our central infrastructure - Working on introducing ElasticSearch

[email protected] ENOG15 | 2018 !30

Anchor VMs

• An anchor is a probe and a willing target - Automatically measured and generates more credits

• Anchor as a Virtual Machine: almost the same - They can be installed where physical machines are a no-go

- Can be “in the cloud” as well

• We’re in the pilot phase as of now - Five VMs are up and running

- With help from members of the community

• We’ll report on this activity around end of Q2

[email protected] ENOG15 | 2018 !31

Probes

• We stopped preparing new v3 probes - Remaining stock is being distributed

• We’re testing v4 (NanoPi based) - Have a few to give out for field testing

- Working on logistics and proper casing

• Still (always?) on the lookout for new devices

[email protected] ENOG15 | 2018 !32

Measurement Tagging

• It allows you to: - Group together any number of your measurements

- Aggregate results from multiple targets and time periods

- Easily stop all of a campaign's ongoing measurements

- Choice between collaboration (tags) and full control (labels)

- Upcoming: support in various RIPE Atlas visualisations

• For more info: - https://labs.ripe.net/Members/chris_amin/ripe-atlas-

measurement-tagging

[email protected] ENOG15 | 2018 !33

References & Feedback

• RIPE Atlas • https://atlas.ripe.net

• Peer-to-peer sketches • http://sg-pub.ripe.net/ixp-country-jedi/

• RIPE IPmap • https://ipmap.ripe.net/

• RIPE Labs • https://labs.ripe.net

Questions